Switch to TSMC for more speed, less power

As the darkness of night falls over China tonight, Qualcomm is hosting a mobile-focused product launch event they’re calling “Snapdragon Night.” Headlined for the event is the announcement of the company’s new flagship SoC, the Snapdragon 8+ Gen 1. A mid-generation update of its flagship smartphone SoC, the Snapdragon 8 Gen 1, the 8+ Gen 1 follows the annual tradition by Qualcomm to release an update product to increase performance and offer something new to partners for the second half of the year. And for this year in particular, we’re looking at a very notable change in Qualcomm’s chips.

Unlike previous generations where Qualcomm merely rolled out a faster speed bin of their existing silicon, we have something more substantial to talk about for 2022. Qualcomm has completely switched the foundries – from Samsung to TSMC – and is launching a new cube as a result. Thanks to this, Qualcomm’s Snapdragon 8+ Gen 1 achieves a one-time manufacturing gain that allows them to increase both CPU and GPU performance while reducing power consumption.

The new Snapdragon 8+ Gen 1 is essentially the original Snapdragon 8 Gen 1, ported from Samsung’s 4nm line to one of TSMC’s 4nm lines. Under normal circumstances, such a shift would likely be inconspicuous – or at most an amusing exercise in edge case hunting – but for Qualcomm’s flagship SoC the matter matters more.

While there are few official sources and statements on the quality of Samsung’s 4nm process, it has become unofficially clear that Samsung’s 4nm process has not lived up to expectations. This has led to cascading effects on chips manufactured on the process node, resulting in the original Snapdragon 8 Gen 1 developing an affinity for power consumption and Samsung’s own Exynos 2200 not performing any better. Conversely, TSMC’s N4 process looks great by all accounts, as the optically downsized node builds on TSMC’s already successful and very capable 5nm technologies.

Because of this performance gap between Samsung’s and TSMC’s 4nm nodes, Qualcomm is taking the unusual step of porting its high-end SoC (essentially) to TSMC’s fab. Which, while not strictly necessary – Qualcomm has plenty of momentum and the 8 Gen 1 has been selling well – is certainly a prudent move for the company. Qualcomm faces particularly strong competition this generation from MediaTek, whose flagship SoC Dimensity 9000 was the main product for TSMC’s 4nm node. And with that, MediaTek has a distinct advantage over the original 8 Gen 1 that Qualcomm would be very keen to squash.

Ultimately, the fab switch gives Qualcomm an opportunity to improve on the original 8 Gen 1 at both ends of the spectrum, resulting in the Snapdragon 8+ Gen 1. On the performance front, TSMC’s node gives you an easy way to increase CPU and GPU clock speeds for more performance. The Prime Cortex-X2 core is now clocked 7% higher at 3.2GHz, and the A710 and A510 clusters, meanwhile, have clocked up even more noticeably by around 12% each. Now even the slowest A510 cores can run at 2GHz. GPU clock speeds have also been increased in a similar fashion, and while Qualcomm doesn’t disclose specific clock speeds there, they have confirmed that the 8+ Gen 1’s Adreno GPU block is clocked 10% higher than the original 8 Gen 1.

But, if anything, most of Qualcomm’s gains from converting manufacturing nodes will be invested in lowering power consumption. TSMC’s better 4nm process is a sore point with the 8 Gen 1 and means Qualcomm is seeing much lower power consumption across its SoC at isofrequency.

Qualcomm officially claims a 30% improvement in GPU and CPU power efficiency. Although as mentioned this is at isofrequency and doesn’t take into account the higher peak clock speeds of the 8+ Gen 1. As a result, the real power savings will not be quite as large at peak times. on a peak basis, but according to Qualcomm the power savings are still significant. Overall, the company is touting a 15% reduction in SoC power consumption under “real-world usage patterns” compared to the original 8 Gen 1, which in turn should result in improved battery life in phones using the new SoC.

Additionally, the official specs for the 8+ Gen 1 do not reflect any significant changes to the SoC’s configuration from the original chip. So we’re still looking at the same integrated X65 5G modem, same Spectra ISP, and same video encoder/decoder blocks (sorry gang, still no AV1 support!). Despite assembling a new chip for the mid-gen Refresh product this year, there aren’t any notable new features on the 8+ Gen 1.

As for SoC performance, Qualcomm officially claims 10% improved GPU and CPU performance due to the aforementioned clock speed increases. While we weren’t able to attend a benchmarking session Qualcomm held last week, the performance mode numbers the company released are roughly in line with those claims. Qualcomm’s Geekbench 5 scores are several percent higher than what we measured at the December 8 Gen 1 launch event, although it’s worth noting that they didn’t score significantly higher in PCMark. GPU performance numbers are similarly mixed, with some of Qualcomm’s official results coming very close to our original 8 Gen 1 results, but I’m hesitant to read too much into them given how much of a difference there is between peak and endurance testing in the GPU results. As always, independent third-party testing will have the final say, although nothing Qualcomm believes at face value is unreasonable given the clock speed improvements and thermal headroom gained by switching to TSMC.

Finally, consumers can get their first glimpse of Snapdragon 8+ Gen 1 devices in the third quarter of this year. According to Qualcomm, many of the usual suspects have committed to releasing phones based on the new SoC, including Asus, Motorola, OnePlus, Honor, and Xiaomi.